We are happy to announce the initial release of the Object Analytics packages. These packages aim to provide real-time object analyses over RGB-D camera inputs, enabling ROS developer to easily create amazing robotics advanced features, like intelligent collision avoidance and semantic SLAM. It consumes sensor_msgs::PointClould2 data delivered by RGB-D camera, publishing topics on object detections, object tracking, and object localization in 3D camera coordination system. Packages and features provided by below 4 projects with tag v0.3.0:
o object_analytics_nodelet: ROS publisher and tests for object tracking in a ROS Image message, and for object localization 3D in a ROS PointCloud2 message from an RGB-D camera. OpenCV tracker is adopted for object tracking, with regular rectification from object detection results. Object segmentation over pointcloud2 image locates the detected objects in 3D camera coordination system.
o object_analytics_msgs: ROS message interface for the object analytics features
o object_analytics_launch: ROS launch interface for the overall object analytics and its nodelet sub-modules
o object_analytics_visualization: ROS sample app with visual outputs for the results of object tracking and object localization 3D
o opencl_caffe: ROS service, publisher, and tests for object inference in a ROS Image message from an RGB camera. Yolo2 model fitting into OpenCL Caffe is adopted for object detection
o opencl_caffe_viewer: ROS sample app with visual outputs for the results of object detection
- object_msgs: ROS message interface common for all object detection packages of various detection framework and hardware
- ros_object_detect_launch: ROS launch interface common for all object detection packages of various detection framework and hardware
We have tested on RealSense D400 series camera, Astra camera, and Microsoft Kinect v1 camera. See README of each project for how to setup the environment. Welcome feedbacks and participation.